Organizational intelligence
We translate leadership judgment into adaptive AI operating systems, so decisions get made your way, at the speed your business actually moves, whether or not you're in the room.
Now booking three founding clients
Who it's for
Companies of 20 to 500 people where a handful of leaders make the calls everything else waits on. Below is a working model of one. The line of light is the signal: your judgment, reaching every handoff.
01 · Chart
We trace real work end to end and find where it waits for a person. Every stall gets a root cause and an annual dollar cost, so the first build goes where the money is.
02 · Bearing
We write down how your leader decides: priorities, trade-offs, dollar limits, voice. Then we send it down the line. Every agent acts on it, and anything above the line goes to a person.
Watch the signal reach each handoff, and the queues clear.
03 · Watch
Every month we test new models against cases from your own work. When a challenger wins on accuracy, speed or cost, we switch it in. Your judgment stays fixed; only the engine changes.
You'll know it's time
Start with Bearing
Start with Bearing
Start with a Survey
Start with a Survey
Start with Field Week
Start with Watch
The Survey, on your numbers
Every time work waits for a person to pick it up, you pay twice: in hours, and in the lead or customer who didn't wait. Put in your own numbers.
Hours lost to handoffs
672per month
people × handoffs × minutes × 21 working days ÷ 60
Get these numbers measured→Example inputs. The Survey replaces them with measured ones.
Why this, why now
10×
Cheaper every year
“For an LLM of equivalent performance, the cost is decreasing by 10x every year.”
A system still running on last year's model is overpaying. Watch is how you collect the difference.
100×
More likely to reach a lead
Calling a new lead within five minutes, rather than thirty, made contact about 100 times more likely and qualification about 21 times more likely.
Speed only pays when the fast answer is the right one. That's what your Bearing is for.
Doctrine
Most AI projects fail on discipline, not technology. These rules hold on every engagement, and they go into every statement of work.
If we don't know how your leader decides, we don't automate the decision.
There is nothing artificial about your judgment. We give it reach, speed and memory.
Field Week ends with the system working on this week's real cases. No pilots that live in slides.
We sit with the people doing the work, inside your systems, until it runs.
Your authority limits are written down. Above them, a person decides, with the facts laid out.
Accounts, keys, code, prompts and data sit in your name from day one. Leave any time and keep it all.
Every automation has an owner, a baseline and a target, and every decision is logged.
What we report
10 of 12
Your calls, matched
Before launch, the system must make your leader's call on at least 10 of 12 real scenarios, and never give an answer they marked as unacceptable.
Fewer
Escalations reaching you
Trust is measured, not assumed. As the system proves itself, fewer decisions need you, with no rise in errors. We report it every month.
Hours
Still counted, just not first
Hours returned, cycle time, errors and cash, against the baseline from your Survey, net of every cost of running the system.
A Bearing, sampled
Describe one decision you make over and over. We'll draft how it might read in your Bearing. It's a first draft from one sentence. The real one takes about four hours with you, and it's tested on twelve real cases before anything runs on it.
Example · a discount to keep a customer
Anything over [your limit], any change to contract terms, or any customer in a dispute comes to you, with the account history laid out.
The question we'd ask you first: what's the dollar amount where you want to be asked?
Write the real one with us →Drafted by AI from one sentence. We don't keep what you type. Privacy
Engagements
One path, fixed prices, and working software on your data in week three.
$7,500fixed
We map how your organization works, price every friction, and write down how your leader decides. Credited in full toward your install if it starts within 30 days.
from $15,000first workflow
Your Bearing goes live inside your systems, starting with the workflow worth the most. By Friday of Field Week your leader watches it handle this week's real cases.
from $3,500per month
We keep the system intelligent: new cases become tests, new models are tried against them, and your Bearing is refreshed every quarter.
If your first workflow hasn't returned its cost within 90 days of going live, we keep working on it at no charge until it does.
Founding client program. Three spots: 40% off your install in exchange for a published case study and a reference call.
In practice
This is Watch up close. Every workflow has a test set drawn from your real past cases, and a new model has to beat the current one on that set, within your Bearing, before anything changes.
Simulated example with illustrative figures. Real reviews use each client's own cases, thresholds and prices.
The Co-Signal Index
Every month we test newly released models on five everyday operations tasks and publish the results on accuracy, cost and speed. First edition coming soon. Tick the box in the form below to get it.
| Task | What we measure |
|---|---|
| Lead triage | Routing accuracy against human labels · cost per 1,000 leads |
| Document intake | Field-level extraction accuracy · pages per minute |
| Call summaries | Required CRM fields captured · reviewer score |
| Inbox triage | Intent accuracy · share of drafts sent without edits |
| Judgment calls | Agreement with a leader's calibration set · unacceptable answers |
Who you work with
Michael Johnston has spent fifteen years where Co-Signal's clients live: on sales floors, in underwriting and in operations, inside organizations where hundreds of people wait on the judgment of a few. He started on the front line. He went on to run a department of more than 150 and to lead leaders across sales and operations.
What he learned there is the premise of this company. Organizations rarely lack intelligence. They lack a way for a leader's judgment to reach the work when the leader isn't in the room. He built that for his own teams first, then led an entire lending organization's move to AI-assisted work. He writes your Bearing himself, builds alongside your team, and stays on the account after it goes live.
Michael JohnstonFounder · Phoenix, Arizona
FAQ
No. Agencies build what they're asked to build. We start with how your leader decides and build only what serves it. Automation is how your organization's intelligence becomes operational; it isn't the product.
A short written spec of how your leader decides: priorities, trade-offs, dollar limits, red lines and voice. It takes about four hours of their time. Every automation reads from it, so the system escalates where your leader would and sounds the way they do.
Whichever wins on your work. We test frontier and open-weights models against cases from your own history, and we build on the CRM, phone system, inbox and ticketing tools you already run.
You do. Code, prompts, test sets and accounts live in your environment, and everything is documented. If you stop working with us, it keeps running.
We build inside your accounts, use model providers' business terms that exclude training on your data, and keep a person approving anything customer-facing or financial until your Bearing says otherwise.
Field Week, in week three, ends with your first workflow running on your own data. It's typically live in production about eight weeks after the Survey starts.
Contact
Be the leader whose judgment runs the place even when you're not in the room. No deck: tell us which decisions wait on you and where your team's hours go, and we'll tell you what we'd build first and what we'd leave alone.